[SPARK-29381][FOLLOWUP][PYTHON][ML] Add 'private' _XXXParams classes for classification & regression

[huaxingao]

What changes were proposed in this pull request?

Add private _XXXParams classes for classification & regression

Why are the changes needed?

To keep parity between scala and python

Does this PR introduce any user-facing change?

Yes. Add gettters/setters for the following Model classes

LinearSVCModel:
get/setRegParam
get/setMaxIte
get/setFitIntercept
get/setTol
get/setStandardization
get/setWeightCol
get/setAggregationDepth
get/setThreshold

LogisticRegressionModel:
get/setRegParam
get/setElasticNetParam
get/setMaxIter
get/setFitIntercept
get/setTol
get/setStandardization
get/setWeightCol
get/setAggregationDepth
get/setThreshold

NaiveBayesModel:
get/setWeightCol

LinearRegressionModel:
get/setRegParam
get/setElasticNetParam
get/setMaxIter
get/setTol
get/setFitIntercept
get/setStandardization
get/setWeight
get/setSolver
get/setAggregationDepth
get/setLoss

GeneralizedLinearRegressionModel:
get/setFitIntercept
get/setMaxIter
get/setTol
get/setRegParam
get/setWeightCol
get/setSolver

How was this patch tested?

Add a few doctest

[huaxingao]

it's a little strange to have setThreshold/Thresholds in the XXXParams class, but scala LogisticRegressionParams does this way, so I just do the same to be consistent with scala side.

[srowen]

Are there any actual changes to the API or it's just moving definitions around?

[SparkQA]

Test build #112181 has finished for PR 26142 at commit 30c4b24.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• class _LinearSVCParams(_JavaClassifierParams, HasRegParam, HasMaxIter, HasFitIntercept, HasTol,
• class LinearSVC(JavaClassifier, _LinearSVCParams, JavaMLWritable, JavaMLReadable):
• class LinearSVCModel(JavaClassificationModel, _LinearSVCParams, JavaMLWritable, JavaMLReadable):
• class _LogisticRegressionParams(_JavaProbabilisticClassifierParams, HasRegParam,
• class LogisticRegression(JavaProbabilisticClassifier, _LogisticRegressionParams, JavaMLWritable,
• class LogisticRegressionModel(JavaProbabilisticClassificationModel, _LogisticRegressionParams,
• class _NaiveBayesParams(_JavaPredictorParams, HasWeightCol):
• class NaiveBayes(JavaProbabilisticClassifier, _NaiveBayesParams, HasThresholds, HasWeightCol,
• class NaiveBayesModel(JavaProbabilisticClassificationModel, _NaiveBayesParams, JavaMLWritable,
• class _MultilayerPerceptronParams(_JavaProbabilisticClassifierParams, HasSeed, HasMaxIter,
• class MultilayerPerceptronClassifier(JavaProbabilisticClassifier, _MultilayerPerceptronParams,
• class _LinearRegressionParams(_JavaPredictorParams, HasRegParam, HasElasticNetParam, HasMaxIter,
• class LinearRegression(JavaPredictor, _LinearRegressionParams, JavaMLWritable, JavaMLReadable):
• class LinearRegressionModel(JavaPredictionModel, _LinearRegressionParams, GeneralJavaMLWritable,
• class _IsotonicRegressionParams(HasFeaturesCol, HasLabelCol, HasPredictionCol, HasWeightCol):
• class IsotonicRegression(JavaEstimator, _IsotonicRegressionParams, HasWeightCol,
• class IsotonicRegressionModel(JavaModel, _IsotonicRegressionParams, JavaMLWritable,
• class _GeneralizedLinearRegressionParams(_JavaPredictorParams, HasFitIntercept, HasMaxIter,
• class GeneralizedLinearRegression(JavaPredictor, _GeneralizedLinearRegressionParams,
• class GeneralizedLinearRegressionModel(JavaPredictionModel, _GeneralizedLinearRegressionParams,

[zhengruifeng]

@srowen This PR should add above-mentioned methods. (Some of them are redundant setters, we will remove in following PR)

@huaxingao some conflicts happened

[huaxingao]

@srowen sorry, I somehow didn't see your comment. As @zhengruifeng mentioned, this PR adds setters/getters in LinearSVCModel, LogisticRegressionModel, NaiveBayesModel,
LinearRegressionModel and GeneralizedLinearRegressionModel. The redundant setters will be remove later in the last parity PR.

@zhengruifeng I will rebase and resolve the conflict.

[SparkQA]

Test build #112253 has finished for PR 26142 at commit cdf87c7.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• class _LinearSVCParams(_JavaClassifierParams, HasRegParam, HasMaxIter, HasFitIntercept, HasTol,
• class LinearSVC(JavaClassifier, _LinearSVCParams, JavaMLWritable, JavaMLReadable):
• class LinearSVCModel(JavaClassificationModel, _LinearSVCParams, JavaMLWritable, JavaMLReadable):
• class _LogisticRegressionParams(_JavaProbabilisticClassifierParams, HasRegParam,
• class LogisticRegression(JavaProbabilisticClassifier, _LogisticRegressionParams, JavaMLWritable,
• class LogisticRegressionModel(JavaProbabilisticClassificationModel, _LogisticRegressionParams,
• class _NaiveBayesParams(_JavaPredictorParams, HasWeightCol):
• class NaiveBayes(JavaProbabilisticClassifier, _NaiveBayesParams, HasThresholds, HasWeightCol,
• class NaiveBayesModel(JavaProbabilisticClassificationModel, _NaiveBayesParams, JavaMLWritable,
• class _MultilayerPerceptronParams(_JavaProbabilisticClassifierParams, HasSeed, HasMaxIter,
• class MultilayerPerceptronClassifier(JavaProbabilisticClassifier, _MultilayerPerceptronParams,
• class _LinearRegressionParams(_JavaPredictorParams, HasRegParam, HasElasticNetParam, HasMaxIter,
• class LinearRegression(JavaPredictor, _LinearRegressionParams, JavaMLWritable, JavaMLReadable):
• class LinearRegressionModel(JavaPredictionModel, _LinearRegressionParams, GeneralJavaMLWritable,
• class _IsotonicRegressionParams(HasFeaturesCol, HasLabelCol, HasPredictionCol, HasWeightCol):
• class IsotonicRegression(JavaEstimator, _IsotonicRegressionParams, HasWeightCol,
• class IsotonicRegressionModel(JavaModel, _IsotonicRegressionParams, JavaMLWritable,
• class _GeneralizedLinearRegressionParams(_JavaPredictorParams, HasFitIntercept, HasMaxIter,
• class GeneralizedLinearRegression(JavaPredictor, _GeneralizedLinearRegressionParams,
• class GeneralizedLinearRegressionModel(JavaPredictionModel, _GeneralizedLinearRegressionParams,

[zhengruifeng]

merged to master, thanks all

[huaxingao]

Thanks! @zhengruifeng @srowen